The Role of Hydrogen in ReRAM

Abstract

Previous research on transistor gate oxides reveals a clear link between hydrogen content and oxide breakdown. This has implications for redox-based resistive random access memory (ReRAM) devices, which exploit soft, reversible, dielectric breakdown, as hydrogen is often not considered in modeling or measured experimentally. Here quantitative measurements, corroborated across multiple techniques are reported, that reveal ReRAM devices, whether manufactured in a university setting or research foundry, contain concentrations of hydrogen at levels likely to impact resistance switching behavior. To the knowledge this is the first empirical measurement depth profiling hydrogen concentration through a ReRAM device. Applying a recently-developed Secondary Ion Mass Spectrometry analysis technique enables to measure hydrogen diffusion across the interfaces of SiO_x ReRAM devices as a result of operation. These techniques can be applied to a broad range of devices to further understand ReRAM operation. Careful control of temperatures, precursors, and exposure to ambient during fabrication should limit hydrogen concentration. Additionally, using thin oxynitride or TiO₂ capping layers should prevent diffusion of hydrogen and other contaminants into devices during operation. Applying these principles to ReRAM devices will enable considerable, informed, improvements in performance.

Keywords: ReRAM; ToF‐ERDA; defects; hydrogen; memristor.

Figure 1

Schematic of the samples and the ToF‐ERDA analysis applied at the University of Jyväskylä, α/β = 20.5°. The ion beam analysis is done in a clear region of SiOx, away from the Au/Ti top electrodes to prevent interference. A high energy (MeV) beam of primary ions (green beam) is used to irradiate the sample surface at a glancing angle, α. This elastically forward recoils sample atoms (red beam) at angle β into two‐timing detectors and a gas ionization chamber, which measure the speed and energy of the atoms, respectively. This enables the determination of the element ejected and its original depth within the material.

Figure 2

IBA measurements of two samples with nominally the same composition, with a similar thickness of SiO_x deposited using a)‐c) Sample 1 – ALD and d)‐e) Sample 9 ‐Sputtering, onto Mo bottom electrodes. a,d) RBS spectra for the sample confirm the concentration of heavier elements. b,e) ToF‐ERDA spectra for both samples, used to generate initial depth profiles (dotted lines) in (c,f). For these depth profiles, concentrations of heavier elements are generated from the beam scatter and are not as accurate due to extended tails from measurement artefacts such as multiple scattering. The solid lines in (c,f) show the final sample composition generated by fitting the ToF‐ERDA results from MCERD to RBS measurements, giving a reliable measure of composition for the full range of elements. For both samples, significant concentrations of H, C, and N impurities were measured, along with oxidation of the Mo electrodes.

Figure 3

a) Electroforming IV sweeps for two devices of each type. Neither of the two Sample 1 – ALD devices electroform, even at ‐15 V, as the film is relatively defect‐free. Sample 11‐ Sputtered has identical top and bottom electrodes to sample 1 but with a sputtered SiO_x layer which electroforms at ≈−5 V in both measured devices. Sample 9 – Sputtered has a thinner sputtered SiO_x which is deposited onto a rougher Mo bottom electrode, hence both devices electroformed at much lower voltages of under −2V. b) XPS depth profiles of samples 3, 4, and 5, each with an SiO_x layer deposited onto Pt, Mo, and Ti bottom electrodes, respectively. XPS sputtering proceeds completely through the SiO_x into the electrode as demonstrated by the drop in measured Si concentration to zero for all three samples at ≈600 s. The location of the SiO_x/electrode interface is taken to be at the half mark of the drop in Si concentration, illustrated by the contrast between the two sections. The inert Pt electrode has negligible oxidation, while the more reactive Mo and Ti electrodes are significantly oxidized. The survey spectra at three depths through this profile are shown in Figure S14 (Supporting Information). Adapted from^{[ 41 ]} with permission of the author.

Figure 4

a) SIMS depth profiles normalized to the total ion signal for pristine HSQ SiO_x samples before and after annealing revealing a loss of hydrogen during annealing. IBA analysis of the annealed sample (sample 6) is shown in Figure S4 (Supporting Information). b) SIMS depth profiles normalized to the total ion signal comparing the hydrogen concentration in sample 4 between pristine devices as well as devices electrically biased with opposite polarities. Increases and decreases in the hydrogen concentration at each depth are illustrated with green and red shading, respectively. This illustrates the reversible field‐driven diffusion of hydrogen to and from the interface under negative and positive voltages applied to the bottom electrode. Corresponding IBA analysis is found in Figure S2 (Supporting Information). Forward and reverse IV sweeps carried out on a comparable SiOx sample to that in b) show: c) electroforming and resetting, d) an initial increase in conduction from cycles 1–5 with a negative bottom electrode bias, e) a decrease in conduction under the same bias from cycles 5–10.